import cv2
import numpy as np
import matplotlib.pyplot as plt


class LBP():
    def __init__(self, img,cell_size):
        self.img = img #灰度圖
        self.height, self.width = img.shape[:2]
        self.cell_size = cell_size
    
    # 傳進(jìn)來一個(gè)int型8位整數(shù) 找到旋轉(zhuǎn)后最小的值
    def find_min(self,code):
        min = code
        for i in range(8):
            code = (code << 1) | (code >> 7)
            if code < min:
                min = code
        return min
            

    def lbp_circle(self):
        # 圓形LBP算子 這里寫成了P=8,R=1 實(shí)際上這個(gè)與矩形3*3的LBP算子是一樣的
        lbp = np.zeros((self.height, self.width), np.uint8) #統(tǒng)計(jì)的lbp是比原來少2行2列的，這里為了方便拆分成cell直接把四周的lbp的值置為0
        for i in range(1, self.height - 1):
            for j in range(1, self.width - 1):
                center = self.img[i, j]
                code = 0
                code |= (self.img[i - 1, j - 1] >= center) << 7
                code |= (self.img[i - 1, j] >= center) << 6
                code |= (self.img[i - 1, j + 1] >= center) << 5
                code |= (self.img[i, j + 1] >= center) << 4
                code |= (self.img[i + 1, j + 1] >= center) << 3
                code |= (self.img[i + 1, j] >= center) << 2
                code |= (self.img[i + 1, j - 1] >= center) << 1
                code |= (self.img[i, j - 1] >= center) << 0
                lbp[i, j] = code
        return lbp
    
    # 旋轉(zhuǎn)不變模式
    def lbp_uniform(self):
        lbp = self.lbp_circle()
        # 圓形LBP算子 這里寫成了P=8,R=1 實(shí)際上這個(gè)與矩形3*3的LBP算子是一樣的
        lbp_uniform= np.zeros((self.height, self.width), np.uint8) #統(tǒng)計(jì)的lbp是比原來少2行2列的，這里為了方便拆分成cell直接把四周的lbp的值置為0
        # 旋轉(zhuǎn)不變模式 找到旋轉(zhuǎn)中的最小值
        for i in range(1, self.height - 1):
            for j in range(1, self.width - 1):
                lbp_uniform[i,j] = self.find_min(lbp[i, j])
        return lbp_uniform
    
    # 等價(jià)模式 這里將等價(jià)的模式按照大小映射到0-57中，然后將不等價(jià)模式標(biāo)記為58
    def lbp_equivalent(self):
        # 定義等價(jià)模式的字典
        uniform_map = {0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 6: 5, 7: 6, 8: 7, 12: 8,14: 9, 15: 10, 16: 11, 24: 12, 28: 13, 30: 14, 31: 15, 32: 16, 48: 17,
                         56: 18, 60: 19, 62: 20, 63: 21, 64: 22, 96: 23, 112: 24,120: 25, 124: 26, 126: 27, 127: 28, 128: 29, 129: 30, 131: 31, 135: 32,143: 33,
                         159: 34, 191: 35, 192: 36, 193: 37, 195: 38, 199: 39, 207: 40,223: 41, 224: 42, 225: 43, 227: 44, 231: 45, 239: 46, 240: 47, 241: 48,
                         243: 49, 247: 50, 248: 51, 249: 52, 251: 53, 252: 54, 253: 55, 254: 56,255: 57}
        lbp_equivalent = np.zeros((self.height, self.width), np.uint8) #統(tǒng)計(jì)的lbp是比原來少2行2列的，這里為了方便拆分成cell直接把四周的lbp的值置為0
        lbp = self.lbp_circle()
        for i in range(1, self.height - 1):
            for j in range(1, self.width - 1):
                if lbp[i, j] in uniform_map:
                    # 按照字典序賦值
                    lbp_equivalent[i, j] = uniform_map[lbp[i, j]]
                else:
                    lbp_equivalent[i, j] = 58
        return lbp_equivalent
    
    # 統(tǒng)計(jì)每個(gè)cell的直方圖 并對(duì)其進(jìn)行歸一化 這里使用的是L2范數(shù)歸一化
    def lbp_histogram(self,lbp):
        bin_size = lbp.max() + 1 # 獲取直方圖的維數(shù)
        # 統(tǒng)計(jì)每個(gè)cell的直方圖
        cell_histogram = np.zeros((self.height // self.cell_size, self.width // self.cell_size, bin_size), np.float32) 
        for i in range(self.height // self.cell_size):
            for j in range(self.width // self.cell_size):
                cell = lbp[i * self.cell_size:(i + 1) * self.cell_size, j * self.cell_size:(j + 1) * self.cell_size]
                cell_histogram[i, j] = np.bincount(cell.flatten(), minlength=bin_size)
              # 對(duì)每個(gè)cell的直方圖進(jìn)行歸一化 這里使用的是L2范數(shù)歸一化
        for i in range(self.height // self.cell_size):
            for j in range(self.width // self.cell_size):
                cell_histogram[i, j] = cell_histogram[i, j] / np.linalg.norm(cell_histogram[i, j])
        return cell_histogram
   
    # 拼接每個(gè)cell，得到最后的LBP特征向量
    def lbp_feature(self,lbp):
        cell_histogram = self.lbp_histogram(lbp)
        # 拼接每個(gè)cell，得到最后的LBP特征向量 是一個(gè)一維向量
        lbp_feature = np.zeros((cell_histogram.shape[0] * cell_histogram.shape[1] * cell_histogram.shape[2]), np.float32)
        for i in range(cell_histogram.shape[0]):
            for j in range(cell_histogram.shape[1]):
                for k in range(cell_histogram.shape[2]):
                    lbp_feature[i * cell_histogram.shape[1] * cell_histogram.shape[2] + j * cell_histogram.shape[2] + k] = cell_histogram[i, j, k]
        return lbp_feature
        
if __name__ == '__main__':
    img = cv2.imread('1.png')
    # 灰度圖讀取
    img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    lbp = LBP(img,16) #這里設(shè)置cell大小為16*16
    lbp_circle = lbp.lbp_circle()
    lbp_uniform = lbp.lbp_uniform()
    lbp_equivalent = lbp.lbp_equivalent()
    print("圓形LBP特征向量:",lbp.lbp_feature(lbp_circle))  # 圓形LBP特征向量
    print("旋轉(zhuǎn)不變LBP特征向量:",lbp.lbp_feature(lbp_uniform))  # 旋轉(zhuǎn)不變LBP特征向量
    print("等價(jià)LBP特征向量:",lbp.lbp_feature(lbp_equivalent))  # 等價(jià)LBP特征向量

     # 顯示原圖和 lbp
    img = cv2.imread('1.png')
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)  # 將 BGR 轉(zhuǎn)換為 RGB 以正確顯示
    plt.subplot(2, 2, 1)
    plt.imshow(img)
    plt.title('Original Image')
    plt.subplot(2, 2, 2)
    plt.imshow(lbp_circle, cmap='gray',vmin=0, vmax=255)
    plt.title('LBP Circle')
    plt.subplot(2, 2, 3)
    plt.imshow(lbp_uniform, cmap='gray',vmin=0, vmax=255)
    plt.title('LBP Uniform')
    plt.subplot(2, 2, 4)
    plt.imshow(lbp_equivalent, cmap='gray',vmin=0, vmax=255)
    plt.title('LBP Equivalent')

    # 保存圖像
    plt.savefig('1_lbp_result.png')

    plt.show()