Scalable Multiresolution Image Segmentation and Its Application in Video Object Extraction Algorithm

Fardin Akhlaghian Tab, Golshah Naghdy, Alfred Mertins


This paper presents a novel multiresolution image segmentation method based on the discrete wavelet transform and Markov Random Field (MRF) modelling. A major contribution of this work is to add spatial scalability to the segmentation algorithm producing the same segmentation pattern at different resolutions. This property makes it suitable for the scalable object-based wavelet coding. The correlation between different resolutions of pyramid is considered by a multiresolution analysis which is incorporated into the objective function of the MRF segmentation algorithm. Allowing for smoothness terms in the objective function at different resolutions improves border smoothness and creates visually more pleasing objects/regions, particularly at lower resolutions where downsampling distortions are more visible. Application of the spatial segmentation in video segmentation, compared to traditional linage/video object extraction algorithms, produces more visually pleasing shape masks at different resolutions which is applicable for object-based video wavelet coding. Moreover it allows for larger motion, better noise tolerance and less computational complexity. In addition to spatial scalability, the proposed algorithm outperforms the standard image/video segmentation algorithms, in both objective and subjective tests.

Original languageEnglish
Title of host publication TENCON 2005 - 2005 IEEE Region 10 Conference
Number of pages6
Publication date31.05.2007
Article number4085093
ISBN (Print) 0-7803-9312-0
ISBN (Electronic)0-7803-9311-2
Publication statusPublished - 31.05.2007
EventTENCON 2005 - 2005 IEEE Region 10 Conference - Melbourne, Australia
Duration: 21.11.200524.11.2005
Conference number: 69661


Dive into the research topics of 'Scalable Multiresolution Image Segmentation and Its Application in Video Object Extraction Algorithm'. Together they form a unique fingerprint.

Cite this