J.-W. Nahm and M.J.T. Smith, "Very Low Bit Rate Data Compression Using a Quality Measure Based on Target Detection Performance," Proc. SPIE, 1995.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....with low (visual) distortion, which is desirable in classification systems requiring visual authentication. 1. INTRODUCTION There exists a variety of applications in which image classification algorithms operate on compressed image data. Examples include Automatic Target Recognition [1, 2, 3, 4], detection of abnormalities in compressed medical images, and classification of compressed aerial images [5] Recently there have been attempts to design the compression algorithm to optimize measures of classification performance. In [5] vector quantization (VQ) schemes are designed so as to ....

....on Cherno# distance as our design criterion. More generally, MSE or a weighted combination of MSE and Cherno# distance can be optimized using a similar technique. There the first term of the cost function in (5) is replaced by the linear combination (1 #)#k# k # # 2 12 , where # [0, 1] is a weighting factor. Note that # = 1 leads to the conventional MSE based design, whereas # = 0 reduces to (5) 5. PERFORMANCE We have outlined the optimal design of a transform coder that minimizes the loss in classification accuracy. We have also presented how a similar procedure can be used ....

J.-W. Nahm and M.J.T. Smith, "Very Low Bit Rate Data Compression Using a Quality Measure Based on Target Detection Performance," Proc. SPIE, 1995.

....optimized transform and multirate coders is compared. The theory is illustrated with a texture classification example. 1. INTRODUCTION There exists a variety of applications in which image classification algorithms operate on compressed image data. Examples include Automatic Target Recognition [1, 2, 3], detection of abnormalities in compressed medical images, and classification of compressed aerial images [4] Recently there have been attempts to design the compression algorithm to optimize measures of classification performance. In [4] vector quantization (VQ) schemes are designed so as to ....

J.-W. Nahm and M.J.T. Smith, "Very Low Bit Rate Data Compression Using a Quality Measure Based on Target Detection Performance," Proc. SPIE, 1995.

....with low (visual) distortion, which is desirable in classification systems requiring visual authentication. 1. INTRODUCTION There exists a variety of applications in which image classification algorithms operate on compressed image data. Examples include Automatic Target Recognition [1, 2, 3, 4], detection of abnormalities in compressed medical images, and classification of compressed aerial images [5] Recently there have been attempts to design the compression algorithm to optimize measures of classification performance. In [5] vector quantization (VQ) schemes are designed so as to ....

....on Cherno# distance as our design criterion. More generally, MSE or a weighted combination of MSE and Cherno# distance can be optimized using a similar technique. There the first term of the cost function in (5) is replaced by the linear combination (1 #)#k# 2 k # # 2 k 12 , where # # [0, 1] is a weighting factor. Note that # = 1 leads to the conventional MSE based design, whereas # = 0 reduces to (5) 5. PERFORMANCE We have outlined the optimal design of a transform coder that minimizes the loss in classification accuracy. We have also presented how a similar procedure can be used ....

J.-W. Nahm and M.J.T. Smith, "Very Low Bit Rate Data Compression Using a Quality Measure Based on Target Detection Performance," Proc. SPIE, 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC