一种改进的图像表示算法

An Improved Image Representation Algorithm

自从极谐波变换(Polar Harmonic Transforms,PHTs)提出后,因其内核计算简单,被广泛用于图像分析和模式识别,而PHTs在高阶时表现出不稳定性,极谐波复指数变换(PCET)在低阶时就有可能表现出不稳定性。为了设计出可以广泛应用于图像表示的矩,提出一种使用新的径向内核函数的极谐波正—余弦变换(Polar Cosine-Sine Transform,PCST)来实现可靠的稳定性。实验结果表明,采用该算法进行图像表示时的稳定性明显优于PHTs,尤其在高阶矩时仍能够保持数值稳定性,同时具有旋转不变性。

Since the introduction of polar harmonic transforms,it has been widely used in image analysis and pattern recognition because it is simple in calculation. However,PHTs shows instability in higher-order transformation and harmonic complex exponential transform（ PCET） may exhibit instability in the low order transformation. In order to design the moment which can be widely used in image representation,a new Cosine-Sine Transform Polar（ PCST） is proposed,which uses a new radial kernel function to achieve reliable stability. The experimental results show that the stability of image representation is better than that of PHTs,especially in the higher order moment. And it has the rotation invariance,and keeps the stability of the image.