SIFT—Python Code

SIFT Pthon-Opencv 源码（待补）

主程序：

def computeKeypointsAndDescriptors(image, sigma=1.6, num_intervals=3, assumed_blur=0.5, image_border_width=5):
    """Compute SIFT keypoints and descriptors for an input image
    """
    image = image.astype('float32')
    base_image = generateBaseImage(image, sigma, assumed_blur) # 金字塔基础图
    num_octaves = computeNumberOfOctaves(base_image.shape)  # 计算塔组数（Octave）
    gaussian_kernels = generateGaussianKernels(sigma, num_intervals)  # 高斯核（卷积）
    gaussian_images = generateGaussianImages(base_image, num_octaves, gaussian_kernels) # 生成高斯pyramid
    dog_images = generateDoGImages(gaussian_images) # 生成DOG 
    keypoints = findScaleSpaceExtrema(gaussian_images, dog_images, num_intervals, sigma, image_border_width) # 找极值
    keypoints = removeDuplicateKeypoints(keypoints)  #  
    keypoints = convertKeypointsToInputImageSize(keypoints)  # 
    descriptors = generateDescriptors(keypoints, gaussian_images)  # 生成描述子
    return keypoints, descriptors

1.生成DOG

def generateBaseImage(image, sigma, assumed_blur):
    """Generate base image from input image by upsampling by 2 in both directions and blurring
    """
    logger.debug('Generating base image...')
    image = resize(image, (0, 0), fx=2, fy=2, interpolation=INTER_LINEAR) # 插值放大2倍
    sigma_diff = sqrt(max((sigma ** 2) - ((2 * assumed_blur) ** 2), 0.01)) #  
    return GaussianBlur(image, (0, 0), sigmaX=sigma_diff, sigmaY=sigma_diff)  # the image blur is now sigma instead of assumed_blur

def computeNumberOfOctaves(image_shape):
    """Compute number of octaves in image pyramid as function of base image shape (OpenCV default)
    """
    return int(round(log(min(image_shape)) / log(2) - 1))

def generateGaussianKernels(sigma, num_intervals):
    """Generate list of gaussian kernels at which to blur the input image. Default values of sigma, intervals, and octaves follow section 3 of Lowe's paper.
    """
    logger.debug('Generating scales...')
    num_images_per_octave = num_intervals + 3
    k = 2 ** (1. / num_intervals)
    gaussian_kernels = zeros(num_images_per_octave)  # scale of gaussian blur necessary to go from one blur scale to the next within an octave
    gaussian_kernels[0] = sigma

    for image_index in range(1, num_images_per_octave):
        sigma_previous = (k ** (image_index - 1)) * sigma
        sigma_total = k * sigma_previous
        gaussian_kernels[image_index] = sqrt(sigma_total ** 2 - sigma_previous ** 2)
    return gaussian_kernels

def generateGaussianImages(image, num_octaves, gaussian_kernels):
    """Generate scale-space pyramid of Gaussian images
    """
    logger.debug('Generating Gaussian images...')
    gaussian_images = []

    for octave_index in range(num_octaves):
        gaussian_images_in_octave = []
        gaussian_images_in_octave.append(image)  # first image in octave already has the correct blur
        for gaussian_kernel in gaussian_kernels[1:]:
            image = GaussianBlur(image, (0, 0), sigmaX=gaussian_kernel, sigmaY=gaussian_kernel)
            gaussian_images_in_octave.append(image)
        gaussian_images.append(gaussian_images_in_octave)
        octave_base = gaussian_images_in_octave[-3]
        image = resize(octave_base, (int(octave_base.shape[1] / 2), int(octave_base.shape[0] / 2)), interpolation=INTER_NEAREST)
    return array(gaussian_images, dtype=object)

def generateDoGImages(gaussian_images):
    """Generate Difference-of-Gaussians image pyramid
    """
    logger.debug('Generating Difference-of-Gaussian images...')
    dog_images = []

    for gaussian_images_in_octave in gaussian_images:
        dog_images_in_octave = []
        for first_image, second_image in zip(gaussian_images_in_octave, gaussian_images_in_octave[1:]):
            dog_images_in_octave.append(subtract(second_image, first_image))  # ordinary subtraction will not work because the images are unsigned integers
        dog_images.append(dog_images_in_octave)
    return array(dog_images, dtype=object)