Your SlideShare is downloading. ×

Thanks for flagging this SlideShare!

Oops! An error has occurred.


Introducing the official SlideShare app

Stunning, full-screen experience for iPhone and Android

Text the download link to your phone

Standard text messaging rates apply

Using Compression Techniques to Streamline Image and Video Storage and Retrieval


Published on

A guide to the use of compression in storing, retrieving and transmitting digital media content, as divergent uses and business purposes mean variable intra-frame and inter-frame image compression …

A guide to the use of compression in storing, retrieving and transmitting digital media content, as divergent uses and business purposes mean variable intra-frame and inter-frame image compression techniques are called for.

Published in: Technology, Art & Photos

  • Be the first to comment

  • Be the first to like this

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

No notes for slide


  • 1. Using Compression Techniques to Streamline Image and Video Storage and Retrieval To overcome large-scale digital media content challenges, organizations need a compression strategy that balances storage and transmission costs with image quality and business requirements. Executive Summary Digital devices such as smartphones and tablets have changed the way in which we live and interact. This channel has paved the way to digitize virtually all work and play operations and activi- ties. Companies across industries now digitize operational and process-oriented content, image and video form, in applications ranging from sur- veillance systems to identity management. The impact is already being felt in industries as diverse as media and entertainment, online retail and social networking. With the rapid develop- ment in mobile technology, users now want quality digital content to be delivered to these new digital channels. Accommodating these demands means delivering images and videos on a large scale in a timely way. This white paper illustrates the challenges that all industry players face in handling large-scale digital media content. Importantly, it proposes a generic solution architecture to help organiza- tions industry-wide to achieve efficient digital media storage and transmission. Digital Content Challenges To meet customers’ increasingly digital content demands, organizations need more cost-effec- tive and secure ways to store, retrieve and share assets. The cost of warehousing digital content increases exponentially with the size/scale of the organization. As such, bigger companies incur greater costs to warehouse digital content. The Bottom Line The storage and transmission of digital images and video has proven to be very expensive. Figure 1 shows the size of a standard definition (720 X 480 X 24bit) video of one hour’s length. For example, a standard definition video of 100GB+ costs roughly $252 per year to store and transmit via a cloud service such as Windows Azure.1 The cost could run into hundreds of thousands of dollars when hours of videos or thousands of uncompressed images in terabytes are stored. The transmission cost and time of such uncompressed data is additional overhead, since it requires gigabytes of bandwidth. In the case of high definition (1920 X 1080) video, the • Cognizant 20-20 Insights cognizant 20-20 insights | may 2014
  • 2. 2 uncompressed file size can be multiplied nearly five times and becomes too large to cost-effec- tively store and transmit. Here is where compres- sion is required. For example, 112GB of video can be reduced to 50GB or less based on the exact components of an effective compression strategy. The Need for Effective Storage Digital content handled across industry sectors such as for surveillance systems is solely for reference. Therefore, it does not demand high- definition output for storage or retrieval. This means organizations can compromise content quality for storage. However, this is not the case in other sectors, such as healthcare. These organizations deal with radiographic images and surgery videos. Content quality in these applications cannot be compro- mised as it concerns critical data. Such sectors demand cost-effective but high-quality data storage and retrieval. Storage solutions in the media and entertain- ment industry require the best of both worlds. It may not be appropriate to broadcast the same quality content over a wide range of devices, because certain digital media content is created with specific devices in mind ranging from smart- phones to high-definition televisions. To under- stand this balance, the quality of the content delivered should be based on the following: • Bandwidth available at the target device. • Maximum quality that the device supports. Therefore, applications that deliver targeted digital content require variable quality retrieval procedures from the same stored media content. The Image Compression Approach Strictly defined, video is comprised of a sequence of images flowing at a fast transmission rate (as measured in number of frames/images per second). Therefore, both image and video com- pression comes down to precision image com- pression. However, video compression follows an additional routine that includes inter-frame com- pression, motion estimation and motion compen- sation. Organizations, therefore, must choose the most appropriate approach for video and image to achieve a desired compression rate. Cost-effective media content storage can be accomplished by using efficient image compres- sion techniques. As mentioned earlier, divergent usages may require different image compres- sion techniques. The level of compression (size reduction) also depends on the specific business need. A counterpart operation of the coding (com- pression) executed at the transmitter side is decoding (decompression), which is performed at the retrieval/receiver end as a part of the image compression procedure. Depending on the media device, user need, transmission bandwidth and its application, the decoder may vary the quality of the digital media content. Two Approaches As mentioned above, a video stream can be viewed as a sequential image stream. Therefore, video coding can be accomplished through image coding with additional processes like inter-frame compression, motion estimation and motion compensation. In image codec terminology, an image is known as a frame. The approach to com- pressing a frame or a sequence of frames can be viewed through two windows: • Intra-frame compression: This technique considers each frame as a non-correlated seg- ment of an image sequence and reduces only the spatial (pixel) redundancies present in an image. • Inter-frame compression: This technique considers each frame as part of an image se- quence and employs temporal predictions, cognizant 20-20 insights Digital Content’s Scale and Scope 1.120E11 Bytes (OR) 112GB 3,600 Seconds 3 Bytes/Pixel 720 X 480 Frame Resolution 30 Frames/Second Figure 1
  • 3. 3cognizant 20-20 insights thus aiming to reduce temporal and spatial (pixel) redundancies. This also increases the efficiency of data compression. Video compression algorithms generally aim for an inter-frame compression technique, because a video stream may have high-ratio temporal redundancies, while an image compression routine will apply an intra-frame compression technique. In addition to inter-frame compres- sion, video coding also follows motion estimation and compensation procedures that are explained in the following sections. Types of Compression Regardless of whether it is image or video, the compression type is broadly classified into two classes: lossy or lossless. The appropriate type of compression that is followed is based on the nature of the application. • Lossless compression: Applications that mandate zero loss in the quality of images and videos upon archiving require the lossless compression technique. Examples are found in healthcare industries which deal with radio- graphic images and manufacturing industries which use machine drawings images and whose intricate details are significant. Similarly, images of circuit diagrams, etc. are another example that demands zero loss in quality and hence use the lossless compression technique. • Lossy compression: Applications that do not require high fidelity in image and video quality are typically archived using lossy compression techniques. The acceptable loss in quality is determined by the use case. A common example of lossy compression relates to images and videos captured by digital cameras or mobile phones, in which data from the image sensor is processed to a compressed format of either GIF or JPEG of desired quality. Lossy compression can reduce the size of the digital content from 5% to almost 95%, depending on the business requirement. Though called “lossy,” they are often termed as visually lossless compression. Implementation Details As depicted in Figure 2, each block in the imple- mentation phase applies a discrete algorithm that is contingent upon the application, approach and the type and level of desired compression. Any digital image is viewed merely as a two-dimen- sional matrix in a spatial domain. The compres- sion algorithm transforms the image, or frame, to a different dimension and domain, in which indi- vidual components of the image can be analyzed. Post analysis, redundant image components are quantized and the image matrix is encoded by using lossless or lossy compression techniques. The encoded image stream is converted into com- pressed (encoded) digital bit streams, which are used for transmission or storage. In the event that the system handles a sequence of images or video streams, motion estimation and compensation algorithms come into play. These components analyze the current frame Image and Video Compression Workflow Encoding Output Compressed Frame Variable Length Coding Represents the Bit Stream with Fewer Bits (Such as Huffman) Quantiser Removes Redundant Bits Compression Algorithm (Such as DCT/ Wavelet) Motion Compensation Motion Estimation Frame Memory Input Frame Figure 2
  • 4. cognizant 20-20 insights 4 with that of the previous frame from the memory and calculate the redundant elements that are inappropriate for further processing. For instance, consider a video stream of about one hour from a surveillance system, which pre- dominantly records an empty aisle with sparse passers-by. Such an uncompressed video stream constitutes more than 100 GB of disk space, which can be expensive to store, transmit and handle. Efficient image compression routines help in reducing cost and minimizing the required disk space, which speeds transmission due to lower bandwidth requirements. In our example of a surveillance video of a mostly empty aisle, most of the video stream is comprised of the same content (i.e., empty aisle). Therefore, frames indicating the empty aisle may be considered redundant and can be encoded into a single frame present at multiple instances. The motion estimation and compensation algorithm reads that the frames are repetitive and responds only to the changes that may be present between the frames, ignoring the redundant components. Tangible Benefits of Compression Techniques • Among the prime advantages of image/video compression is size reduction. Based on the application, an image/video stream can be compressed to the required size, which can eventually save storage space. Therefore, it is cost-effective. For situations that involve large amounts of data, this generates significant impact, optimizing archiving and producing appreciable cost savings. • Another advantage is the variable quality retrieval procedure. This is also referred to as progressive resolution encoding. Some image/video playback devices do not support high-resolution data streams or may restrict the bandwidth through which the content is delivered. This prevents streaming of uncom- pressed high-quality data over such devices. In such cases, image decoder (decompres- sion) can be tuned to deliver appropriate data quality from the compressed stream. • Transmission of compressed digital media content between electronic devices or Web hosts and the device’s retrieval rate is faster, thereby improving the workflow of the process. Compression Au Courant New modes of image compression techniques are emerging that could reduce costs for many industries such as online retailing and market researchers whose businesses pivot around huge image volumes. Other industries such as media and entertainment could also embrace new compression techniques to achieve greater effi- ciency. Consider the example of the surveillance system. The hours of surveillance video recording and thousands of photographic images for identity management (which are stored for mere reference) occupy large volumes (in terabytes) of server space. Similar scenarios prevail in online commodities businesses, where millions of com- pany-supplied, user-shared product images of products are stored on its e-commerce servers. Market Insight • Apopulargraphically-richiOSTwitterclientapp, Tweetbot, which weighs over 33MB, uses over 26MB to stock more than 900 images2 that are compressed using a native iOS IDE compres- sion technique. Such a large image size slows image display time. Beyond this elementary level compression, the app is integrated with additional compression techniques to achieve a compression rate of more than 80% and a display time that is reduced by a factor of three (see Figure 3). Figure 3 Compression Analytics Lossy Compression Technique Lossless Compression Technique Native Compression — iOS IDE Uncompressed 9.37 MB 16.81 MB 26.46 MB 49.63 MB
  • 5. cognizant 20-20 insights 5 • A popular camera-capturing app on iOS, IncrediBooth, was 70MB in size and had PNG images of screen texture with a resolution of 2048 X 1536, each weighing more than 10MB in its app bundle. The bundle size was shrunk from 70MB to 31MB3 (a reduction of more than 50%) by incorporating effective lossless image compression techniques. • Experimental results in healthcare show that medical imaging (such as Dicom, CT, MRI, etc.) can be optimized using compression schemes that retain image quality only in the region of interest (i.e., in diagnostically important regions) and reduce image size by more than 95%.4 • A Netherlands-based medical imaging systems firm, Accusoft, has implemented lossless JPEG compression and decompression schemes to achieve image quality comparable to the original image. A typical 2048 x 2048 resolution 16-bit grey scale image of size 9,948 KB was reduced to 5,517 KB5 (more than 40% compressed) using lossless compression techniques. Looking Forward When a state-of-the-art image compression strat- egy is embraced, companies can save on storage and transmission costs. Similarly, retrieval rates can also be increased, thereby providing fast and effective solutions to users inside and outside the company. The ratio of the storage savings is directly proportional to the level of compression and the quality required upon retrieval. When it comes to digital content management, every block in the implementation workflow plays a vital role in warehousing it effectively. An effective algorithm is required for each block for efficient repositioning. To achieve cost-effective warehousing and trans- mission of digital media content, organizations must analyze the level of quality/compression required by the business. The resulting solution should then be capable of handling multiple images and video formats as required by the business. The compression techniques selected should be framed and optimized to achieve the required compression rate at the desired data quality, without incurring any additional overhead. Finally, if the business targets multiple devices for content delivery, then the decoder (at the device end) must be modelled to deliver progressive quality based on the device specification. Footnotes 1 2 3 4 P. Bharti, S. Gupta, and R. Bhatia, “Comparative Analysis of Image Compression Techniques: A Case Study on Medical Images in Advances in Recent Technologies in Communication and Computing,” ARTCom ‘09. International Conference, 2009. 5
  • 6. About Cognizant Cognizant (NASDAQ: CTSH) is a leading provider of information technology, consulting, and business process outsourc- ing services, dedicated to helping the world’s leading companies build stronger businesses. Headquartered in Teaneck, New Jersey (U.S.), Cognizant combines a passion for client satisfaction, technology innovation, deep industry and busi- ness process expertise, and a global, collaborative workforce that embodies the future of work. With over 75 develop- ment and delivery centers worldwide and approximately 178,600 employees as of March 31, 2014, Cognizant is a member of the NASDAQ-100, the S&P 500, the Forbes Global 2000, and the Fortune 500 and is ranked among the top performing and fastest growing companies in the world. Visit us online at or follow us on Twitter: Cognizant. World Headquarters 500 Frank W. Burr Blvd. Teaneck, NJ 07666 USA Phone: +1 201 801 0233 Fax: +1 201 801 0243 Toll Free: +1 888 937 3277 Email: European Headquarters 1 Kingdom Street Paddington Central London W2 6BD Phone: +44 (0) 20 7297 7600 Fax: +44 (0) 20 7121 0102 Email: India Operations Headquarters #5/535, Old Mahabalipuram Road Okkiyam Pettai, Thoraipakkam Chennai, 600 096 India Phone: +91 (0) 44 4209 6000 Fax: +91 (0) 44 4209 6060 Email: ­­© Copyright 2014, Cognizant. All rights reserved. No part of this document may be reproduced, stored in a retrieval system, transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the express written permission from Cognizant. The information contained herein is subject to change without notice. All other trademarks mentioned herein are the property of their respective owners. About the Author Srinivasan Krishnan is an Associate with Cognizant’s Audio Video Imaging Center of Excellence within the company’s Global Technology Office. He has over four years of architecture design, algorithm development and application-building experience and has extensive experience developing solutions in niche domains such as digital image processing, computer vision and machine learning. He holds a bachelor’s degree in electronics engineering from Anna University and a master’s in microelectronics systems design from University of Southampton in the UK. He can be reached at