An evolutionary approach for image retrieval based on lateral inhibition

• An evolutionary image retrieval framework is introduced.
• A lateral inhibition based image enhancement procedure is adopted to facilitate image retrieval.
• Whether it is beneficial to adopt evolutionary algorithms has been investigated.
• Whether it is beneficial to apply lateral inhibition has been investigated.
• Convergence analysis of our concerned evolutionary algorithm is provided.

Image retrieval refers to searching for specific pattern when browsing digital images in large databases. It is a fundamental topic in visual pattern recognition. The original image retrieval scheme is converted into a 2-dimensional optimization problem, during which image enhancement, a preprocessing step, is implemented to ease the template matching procedure. Image enhancement is achieved via lateral inhibition. Fundamental issues in this topic are investigated in this paper: (1) whether it is sensible to adopt an evolutionary algorithm to search for the best-match result, given that the concerned optimization problem has merely two dimensions; (2) how lateral inhibition takes effect to facilitate image retrieval. According to our comparative experimental results, there is no evidence that indicates lateral inhibition makes any subtle effort to ease the best-match optimization process. This paper also provides theoretical analyses regarding the evolutionary algorithm adopted.

Figure optionsDownload as PowerPoint slide