Description of evolutional changes in image time sequences using MPEG-7 visual descriptors
Colour and texture visual descriptors have been developed to represent structural features of images, mainly under the Query-by- Example (QBE) image retrieval paradigm. This paper explores applicability of MPEG-7 visual descriptors to describe and measure evolutional changes in image time sequences, using a fruit rotting process as an example. The research found that MPEG-7 visual descriptors can be applied to describe evolutional changes in image time sequences. The experimental results are provided using bananas captured in image time sequences. The results show the desirable monotonicity of description metrics of MPEG-7 similarity matching for image time sequences and their sensitivity to practical image acquisition conditions. Our experiments demonstrate that Colour Layout descriptors (CLD) and Scalable Colour descriptor (SCD) describe the changes that are consistent to the degree of visual changes while CLD and Homogeneous Texture descriptor (HTD) are more robust to variations of image data due to practical image acquisition conditions.