Unstructured Scene Object Localization Algorithm Based on Sparse Overcomplete Representation
Unstructured scene has many uncertainties and unpredictable states. It brings difficulties to the object localization, which is pixel-based processing. Themethod of analog visual information processing is an effective way to solve the problem mentioned above. Sparse overcomplete representation is an image representation model which is more in line with visual mechanism. However, the overcomplete representation not only increases the combinatorial search difficulty of sparse decomposition, but also changes the symmetry between input and code space. Furthermore, it makes the model solution and calculation method complicated. In order to solve the afore mentioned problem and effectively use this model to achieve automatic image object localization, this paper takes the unstructured scenes object localization as the background. Firstly, the overcomplete representation computational model which is based on energy model and score matching method is established. Then an automatic object localization method based on the neuronal response and dynamic threshold strategy is proposed and applied to the movement object localization. On this basis, the error analysis is done. Experimental results show that the method can achieve the movement object localization.
KeywordsUnstructured scenes sparse overcomplete representation object localization score matching
Unable to display preview. Download preview PDF.
- 1.Rodriguez, M., Ali, S., Kanade, T.: Tracking in Unstructured Crowded Scenes. In: IEEE 12th International Conference on Computer Vision, pp. 1389–1396 (2009)Google Scholar
- 2.Lu, P., Li, Y.Q.: Fault-image Detection Algorithm Based on Visual Perception. Chinese Journal of Scientific Instrument 31(9), 1997–2002 (2010)Google Scholar
- 7.Cai, Z.M., Lai, J.Z.: An Overcomplete Learned Dictionary-Based Image Denoising Method. Acta Electronica Sinica 37(2), 347–350 (2009)Google Scholar