Evolutionary Optimisation of JPEG2000 Part 2 Wavelet Packet Structures for Polar Iris Image Compression

Abstract

Abstract. The impact of using evolutionary optimised wavelet subband stuctures as allowed in JPEG2000 Part 2 in polar iris image compression is investigated. The recognition performance of two different feature extraction schemes applied to correspondingly compressed images is compared to the usage of the dyadic decomposition structure of JPEG2000 Part 1 in the compression stage. Recognition performance is significantly improved, provided that the image set used in evolutionary optimisation and actual application is identical. Generalisation to different settings (individuals, sample acquisition conditions, feature extraction techniques) is found to be low.

Effects of severe image compression on iris segmentation performance

“…JPEG Q-table optimisation has already been considered in face recognition [15] which leads to superior recognition performance as compared to the standard matrix. A further example is the optimisation of JPEG 2000 Part 2 wavelet packet decomposition structures with respect to optimising iris recognition accuracy which provides better results compared to ratedistortion optimised wavelet packet structures [9]. These observations raise the question if compression algorithms exhibiting better rate-distortion performance are also better in a biometric recognition context.…”

“…JPEG-XR is found to be competitive to the current standard JPEG 2000 while exhibiting significantly lower computational demands. For JPEG 2000, Hämmerle et al show that JPEG 2000 Part 2 can improve recognition accuracy by employing adapted wavelet packet basis instead of the fixed dyadic Part 1 decomposition [9] and also, JPEG 2000 region-of-interest coding is shown to by beneficial by assigning a higher bitrate to iristexture regions in the image [10].…”

Effects of image compression on ear biometrics

“…Aforementioned studies as well as standardisation activities analyse the effects of different image compression standards on biometric recognition systems in order to elaborate a common standard for the compression of biometric data which still enables reliable authentication [19]. Further, diverse compression techniques have been explicitly designed for distinct biometric modalities, for example, in [20,[35][36][37][38]. Hence, in various application scenarios lossless or slight compression is considered in bi-lateral compression scenarios, that is, probe and reference images…”

Recognition Impact of JPEG2000 Part 2 Wavelet Packet Subband Structures in IREX K3 Iris Image Compression

“…Recent work [10] showed that for polar iris image (IREX K16) common subband structure selection strategies including rate-distortion optimising ones are not very successful as compared to the dyadic decomposition scheme (defined in the Part 1 of the JPEG2000 standard suite). However, we were able to demonstrate some limited performance gain using evolutionary optimization for selecting wavelet packet subband structures [11]. In this work, we re-investigate the usage of common wavelet packet subband structure selection strategies including rate-distortion optimising ones for rectilinear K1/K3 imagery.…”