Adaptive subsampling JPEG image coding

We present a JPEG-based image coding system which preprocesses the image by adaptive subsampling based on the local activity levels. As a result, it yields good quality in both smooth and complex areas of the image at high compression ratios. We propose an algorithm aiming at the improvement of the image quality at higher compression ratios than that JPEG can handle. This scheme combines decimation/interpolation and activity classification as the pre-/post-processing, and uses JPEG with optimal Q-tables as the compression engine. It yields better image quality than the original JPEG or the uniform subsampling JPEG. The increased complexity is only minor compared to JPEG itself