For Project Enquiry +91 9962 067 067

EXISTING SYSTEM

PROPOSED SYSTEM

EXISTING CONCEPT:

          Texture descriptors, one can develop a combined spectral–texture segmentation framework by feeding integrated features to clustering approaches to produce segmentation.

          However, there are two main problems associated with such framework. First, applying multiple filters to spectral bands generates high-dimensional features.

PROPOSED CONCEPT:

          We use local spectral histogram representation, which consists of histograms of filter responses in a local window. This representation provides an effective feature to capture both spectral and texture information.

          This method works across different bands in a computationally efficient way and accurately localizes boundaries.

EXISTING ALGORITHM:

          Texture Descriptors

PROPOSED ALGORITHM:

         Local Spectral Histogram Representation

ALGORITHM DEFINITION:

          Texture descriptors constructed by analyzing the local distribution of filter responses have been shown to be powerful features for texture synthesis and discrimination.

ALGORITHM DEFINITION:

          Local spectral histogram representation which consists of histograms of filter responses in a local window.

DRAWBACKS:

          It is difficult to acquire satellite images with high resolutions in both spectral and spatial domains.

         Texture descriptor is the high computational cost, which makes it impractical for large images.

ADVANTAGES:

          Segmentation based on spectral and texture features. The results can be easily improved by incorporating additional information.

         A more general approach is to impose constraints on a least squares solution, making it more robust to noise.