Objective Image Quality Metrics for DCT-Based Video Compression

Abstract

Image quality assessment on compressed digital video is an indispensable task in tuning compression algorithms for different applications as well as evaluating different encoding algorithms, implementations and encoder/decoder products. Since in most compressed video communication applications the ultimate viewers are human beings, it is desirable that the results of an image quality assessment procedure be consistent with the quality perceived by a typical viewer. The design of truly human visual system (HVS) compatible objective image quality metrics, consequently becomes a task of extreme importance. Block based application of the Discrete Cosine Transform (DCT) became the data compression tool of choice in many international video and image compression standards (MPEG-1, MPEG-2 (H.262), MPEG-4, H.261, H.263, H.263+, H.263++, JPEG) owing to its many desirable features. Nevertheless, the manner block based DCT is applied as well as information quantization in the transform domain lead to undesirable compression artifacts. In this paper, three objective image quality metrics are introduced. These metrics are designed to quantify the severity of the three major types of artifacts associated with block DCT based image/video compression schemes, in a manner compatible with HVS. In a vendor transparent manner, experimental results from some commercially available encoding applications are reported to illustrate the practical usefulness of the proposed metrics to assess compressed image quality aside from their theoretical significance.