Flower image segmentation with PCA fused colored covariance and gabor texture features based level sets

Abstract

This paper presents a framework for segmenting flower images captured with a digital camera. Segmenting flowers from images is a complex problem attributed to translation, scaling, rotation with variable backgrounds in each captured image. We propose to solve this problem using principle component analysis based color texture fusion as a prior parameter for level set evolution (FCTAC). First, Color Gabor textures (CGT) and Color Level Covariance Matrix (CLCM) texture features are extracted. Principle component analysis based fusion constructs a color discriminative texture as a knowledge base with convex energy function for active contours without edges. The proposed global segmentation framework with fused textures will avoid the local minimums during curve evolution. We test the proposed segmentation model on the benchmark oxford flower image dataset and our own dataset. The results of FCTAC were tested against the state-of-the-art methods in accuracy and efficiency.