HOW IMAGE DOWNSCALING AND JPEG COMPRESSION AFFECTS IMAGE CLASSIFICATION PERFORMANCE - An experimental study
Abstract: the quality of an image plays a role in how well it can be correctly classified by an image classifying neural network. Artifacts such as blur and noise reduces classifilability. At the same time it is often motivated to reduce file sizes of images which also tends to introduce artifacts and reduce their quality. This leads to a trade-off between having small file sizes and achieving high classification accuracy. The two main approaches of reducing file sizes of images is to either reduce the number of pixels in them via image scaling or to use less data to represent each pixel via compression.The effects of these two approaches on image classification accuracy have been studied independently.In this study the effects of combining image scaling and compression in regards to image classifiability is examined for the first time. Images are downscaled using five popular methods before being compressed with different magnitudes of JPEG compression. Results are evaluated based on the fraction of the treated images that are correctly classified by the classifier as well as on the image file sizes.The results shows that the scaling method used has significant but weak effect on image classifiability. Thus the choice of scaling method does not seem to be critical in this context. There are however trends suggesting that the Lanczos scaling method created the most classifiable images and that the Gaussian method created the images with highest classiability to file size ratio. Both scaling magnitude and compression magnitude were found to be better predictors of image classifiability than scaling method.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)