An efficient implementation of the channelized Hotelling observer for task-based assessment of lossy compressed images

Abstract

An eMclent suhhand implementation of the channelized Hotelling observer is presented, whieh eau be used to assess the image quality of Images eompressed with waveletbased techniques. The channelized Hotelling model observer has heen shown to prediet human performanee lo detecting signals in noise-limited images. The model observer can also prediet degradation of human performance due to lossy compressed images. This provides a more relevant image quality assessment for medical images, where the image's value Is in supporting clinical decisions, than metrics sueh as mean square error. The suhhand implementation shown is unique in that It operates on ehannel responses of the wavelet suhhands rather than on the entire image itself. The technique is extendable to operate on the ehannel response of the subhand hitmaps, whieh would permit hit ordering optimized for human Derformanee.