Evaluation of Classification Accuracy with Original and Compressed Images

Abstract

The extent of communicated information through internet has augmented speedily over the past few years. Image compression is the preeminent way to lessen the size of the image. JPEG is the one the best technique related to lossy image compression. In this paper a novel JPEG compression algorithm with Fuzzy-Morphology techniques was proposed. The efficacy of the proposed algorithm compared to JPEG is presented with metrics like PSNR, MSE, No of bits transmitted. The proposed approaches lessen the number of encoded bits as a result tumbling the quantity of memory needed. The Planned approaches are best appropriate for the images corrupted with Gaussian, Speckle, Poisson, Salt & Pepper noises. In this paper the effect of compression on classification performance was envisaged, Artificial Neural Network, Support Vector Machine, and, KNN classifiers performance is evaluated with original image data, standard JPEG compressed data and the compressed image data with the proposed method.