Your SlideShare is downloading. ×
Cp31608611
Cp31608611
Cp31608611
Cp31608611
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×
Saving this for later? Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime – even offline.
Text the download link to your phone
Standard text messaging rates apply

Cp31608611

56

Published on

IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com

IJERA (International journal of Engineering Research and Applications) is International online, ... peer reviewed journal. For more detail or submit your article, please visit www.ijera.com

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
56
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
1
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide

Transcript

  • 1. Pratap Khanna Kolli, Karunaiah. B / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.608-611 Implementation of Image Thumbnails Using Resizing Method Pratap Khanna Kolli1, Karunaiah. B2 1 M. Tech Student (Embedded systems), Holy Mary Institute of Technology & Science, Hyderabad 2 Professor ECE, Holy Mary Institute of Technology & Science, HyderabadAbstract A standard image thumbnail is generated modify image composition while resizing. The paperby filtering and sub sampling when the blur and addressing image with different types of contentsnoise of an original image is lost since the such as webpages, documents and photographs. Thisstandard thumbnails do not distinguish between paper develops image previews that address imagehigh quality and low quality originals. In this quality to use iterative gradient field computations topaper an efficient algorithm with a blur – accentuate selected image features.generating component and a noise generating There are many applications for the newcomponent preserves the local blur and the noise thumbnails. A digital camera display could provide aof the originals. The new thumbnails are more quick look at the image quality of the originals. If therepresentative of their originals for blurry images images are of low quality, the user could immediately.The noise generating component improves the take another picture. In addition, current photoresults for noisy images but degrades the results browsing applications and operating systems bothfor textured images .The decision to use the noise provide image thumbnails during browsing.component of the new thumbnails should based on Thumbnails are often generated on embedded devicestesting with the particular image mix expected for with limited computational power, or they arethe application. generated in batch in large numbers. To find wide spread use, quality-preserving thumbnails needKeywords – Standard thumbnails, image quality, efficient computation.noise modeling This paper is organized as follows section IV describes the new algorithm for space-varyingI. INTRODUCTION blur estimation and blur generation for the new Filtering and sub sampling prevents aliasing thumbnails, and section V describes the fast methodand preserves image composition and loses its image for preserving noise in the thumbnails. Section VIquality .Both sharp and blurry original appear sharp describes subjective experiments testing thein the thumbnails, and both clean and noisy originals thumbnails against standard thumbnails .section VIIappear clean in the thumbnails This leads to errors ends finally with the conclusion.and inefficiencies during image selection based onthe thumbnails rather than the high resolution II IMAGE MODEL AND FORMULATIONoriginals. The derivation below use 1-D column-stacked vector The standard thumbnails introduced in [3] notation to simplify the representation as vector X, isdo not modify the image composition but better (1)reflects the image quality of the originals. The Newthumbnails provide a quick, natural way for users to In this equation the vector S represents anidentify images of good quality at the same time that ideal image captured with infinite depth of field. Thethey select images with desired subject matter. The matrix B represents a space-varying blurnew thumbnails are natural to interpret; there is no corresponding to unintended blurs such as cameralearning necessary to understand the blur and noise in shake, motion blur or misfocus as well as intendedthe new thumbnails. The alternate approach of depth of blur and n represents anautomatic image ranking by quality is extremely additive,perhapscorrelated,noise.Digital photographsdifficult because the knowledge about the subjects of taken under ideal conditions have no unintended blurinterest resides with the user, not with the image .In or noise in this case the noise But thethe new thumbnails the user can check whether the matrix B is not necessarily the identity ,since it issubject of interest is in focus. still represents the space-varying blur due to limitedSeveral recent papers offer interesting, nontraditional depth of field together with objects of differentalgorithm for reducing image size. None of the distances. Only in the special case of infinite depth ofalgorithms address image quality and all of them field does B=I and therefore X=S. 608 | P a g e
  • 2. Pratap Khanna Kolli, Karunaiah. B / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.608-611The goal is to recover S from X generates a low [15].The blur map is determined without the type ofresolution representative thumbnail tr ,not the exact identification of the type of the blur.reconstruction of high resolution S.For exaample our The low pass images are created by convolving thesolution works well with both shake and defocus standard thumbnail with Gaussian kernels of differentblurs,by applying an appropriate space varying variances is given asblur.The exact form of the applied blur kernel is notcritical. The new thumbnail is generated by starting (5)first with a standard thumbnail generated by filtering The number of scale space images determines howand subsampling, ts,which is blur and noise free even finely the blur can be represented but the algorithm isfor distorted input images.To this standard insensitive to the number of scales.thumbnail,blur and noise are applied to correspond tothe blur and noise in the original (2)III. NEW COMPANDING ALGORITHMFig.1.To generates the new thumbnail, the standardthumbnail is modified with a space-varying blur andan estimated noise is added.III STANDARD THUMBNAILS Fig.2.Blur Process involves comparing subsampleThe standard thumbnails is given by high resolution pixel ranges in a scale space (2) expansion.Where it combines filtering andsubsampling.Expanding (2) using the image The blur map shown in Fig.1. And Fig.2.ismodelling (1) results in computed by comparing local range images. The generation of the blurred thumbnail images is linear (3) in the number of thumbnail pixels. The bandwidth of the blur is broader thanthe bandwidth of the antialiasing filter for typical IV NOISE COMPUTATIONsubsampling factorsThus in (4) noise next,antialising A. Introduction and problem Formulationfilter applied result in output filtered noise variance The noise generation developed bymuch lower than the input variance.The case of a k X borrowing directly from image denoising, works wellk boxcar filter,for a subsampling factor k is but it slow. Adding random jitter to the sub samplingparticularly easy to analyze.If the input noise is breaks up potential moiré from any residual imageuncorrelated,the output noise variance will be textures that may erroneously appear in the noisereduced by a factor of 1/k^2. image. The total blur and noise computation forThe Mean square error(MSE) between a distortion converting an original image takes 0.14 s instead ofimage and undistorted image is proportional to the the original 2.25 s on a 2-GHZ processor and 1GBsquare of the Eucledian norm between the images Ram.interpreted as vectors.The distorted image may be expressed as the addition The results in image model is given byof the two vectors to the original image ,given by (6) (4) Where x is a noisy originalimage,s is an idealimage.IV. LOCAL BLUR COMPUTATION S=undistorted image and n=noise Many prior methods for image blurdetermination provides global blur metrics that assessthe overall quality of an image. The approach toprovide local blur for detecting edges is given in 609 | P a g e
  • 3. Pratap Khanna Kolli, Karunaiah. B / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.608-611 confirming that the noise MSE in this case is image independent. VII PERFORMANCE SIMULATION The input image is 620x792 pixels and the size is 25.4KB by applying standard thumbnail and the new thumbnails the simulations are taken blur versus Mean square Error (MSE) and blur versus PSNR. At both resolutions, usersfind that the new thumbnails are (1) Better represents the blurry andFig.3.Block diagram for noise preservation based on noisy images (2) They are not significantly differentdenoising. than standard thumbnails for the clean images. For most of the textured images, however, the usersB. Solution Using Denoising and Sub sampling prefer the standard thumbnails.The Threshold is determined by first robustly Thumbnail treatments were positioned toestimation noise standard deviation appear on the right and left side randomly. ImageThe computation of the threshold requires the high samples were presented to each observer in aresolution signal. There are enough pixels even at the different random order. This technique distributesthumbnail resolution, for an effective determination any start up effects over different samples.of the noise variance of the original signal.Fig.4.Block diagram for the noise generationalgorithm. The difficulties of the noise generatingalgorithm are in distinguishing between highfrequency textures from noise. The nonlinear soft clipis memoryless,it commutes withsubsampler.Eventhough there are nonlinear steps inthe original algorithm for determining the thresholdvalue and also for applying the memory lessnonlinear operation.The algorithm was first tested informally with severalimages and it was found effective for blurry andnoisy images. There were however, differences The total elapsed time that is blurring and noise timebetween the standard and new thumbnails for for converting an original image into new thumbnailtextured images. By turning off the noise processing, is 194.760236seconds.corresponding to an additive noise . The complexity of the blur computation is composed of two parts (1) the complexity of The noise term was found to account for the generating the standard thumbnail and extractingdifferences in the textured images. This is due to the features which are both linear in the number of thecurrently used noise algorithm, which does not original image pixels since both computationsalways distinguish between image noise and texture, compromise local computations in small constantboth of which contains high spatial frequencies are in sized neighborhoods.(2) The generation of thethe input image that results in more blur distortion. blurred thumbnail images is linear in the number ofThe faster increase of MSE along the blur axis 610 | P a g e
  • 4. Pratap Khanna Kolli, Karunaiah. B / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com Vol. 3, Issue 1, January -February 2013, pp.608-611thumbnail pixels. Thus the overall complexity is AUTHOR’S BIOGRAPHYlinear in the number of original pixels. Mr.Pratap Khanna Kolli didVIII CONCLUSION B.Tech in Electronics and A new image resizing method that preserves communication Engineering fromthe image blur and the image noise .The subjective SIR C R Reddy College ofevaluation of the thumbnail shows that the blur Engineering ,Affiliated to Andhracomponent of the algorithm is robust and it may University, Visakhapatnam andalways be used with improved results. Equal number Pursuing M.Tech in Embeddedof the different image categories was used to better systems from Holy Marytest the different cases. institute of Technology and Science, JNTU Hyderabad.ACKNOWLEDGEMENTS His interested areas are I grateful thanks to management of, Holy Embedded systems and imageMary Institute of technology and science, for him processingstimulating guidance and generous assistance. Karaunaiah B is pursuing PhDREFERENCES from Andhra University, [1] A.K.Jain Fundamentals of Digital Image Visakhapatnam. In the field of Processing.Englewood cliffs NJ; Prentice- Antennas. Completed M. Tech in Hall,1989. ECE with Electronic and [2] A.Munoz,T.Blu,and M.Unser,”Least- instrumentation from Andhra squares image resizing using finite University, Visakhapatnam.B. Tech differences”,IEEE Trans.Image in Electronics and Communication Process,2001. Engineering from Bapatla [3] R.Samadani,S.Lim,and D.Tretter, Engineering college, Nagarjuna “Representative image thumbnails for good University, Guntur. Currently browsing,”in Proc.IEEE Int.Conf.Image working as Professor in the ECE Processing,sep2007,vol.II. Department at Holy Mary Institute [4] Y.Ke,X.Tang,and F.Jing, “ The design of of Technology and Science (College high-level features for photo quality of Engineering), Hyderabad. Areas assessment,”in IEEE Computer Society of interest are optical Conf.Computer Visionand Pattern communications, em field theory, Recognition,2006,vol.1. radar engineering, microwaves, [5] antenna & wave propagation, A.Wondruff,A.Faulring,R.Roshenholtz,J.Mo electronic devices, satellite and rrison,andP.Prolli, “Using thumbnails to digital communication search the web,”in Proc.CHI,Apr.2001. [6] B.Suh,H.Ling,B.Bederson,and .Jacobs,S.Wink “Automatic thumbnails cropping and its effectiveness,” Proc.16th Annu.ACM,Dec.2003. [7] K.Berkner,E.Schwartz,and .Marle,L.Karamin“SmartNails:Display-and the image-independent thumbnails,” in Proc.SPIE Document Recognition and Retrieval XI,Dec.2003. [8] S.Avidan and A.Shamir, “Seam carving for content-aware image resizing,”ACM Symp.User In Trans.Graphics,vol.26,2007. [9] L.Wan,W.Feng,Z.Lin,T.Wong, andT.Ebriamian Z.Liu, “Perceptual image preview,”,oct.2008. 611 | P a g e

×