A novel method for salient image enhancement

Visual Salient Object Detection and enhancement plays a vital role in many computer vision applications. In the traditional image processing approaches which are used to extract salient maps have various assumptions such as color, contrast, context, gradient and histogram. The existed models have achieved good progress to extract saliency map on selected features, still considerable features are unclear. The detection sensitivity of the previous models to object interior texture, background textures, object boundaries and blurred contours is same. The proposed model is based on multi-scale transforms and it is developed by contrast based enhancement operator. This approach includes properties of multi-scales and entropy measure of image. By using this we can achieve more complete and uniformly enhanced salient object regions. This model can enhance the salient object interior textures, background texture while simultaneously reducing blur on object boundaries. It achieves better enhancement measure of entropy, frequency measure and mean absolute errors which are the performance measures of salient object enhancement while compared to the state of art methods.