```python
# TODO 使用K-means实现一幅纹理图像的滤波响应向量聚类
import numpy as np
from skimage import color
from skimage.filters import gabor
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt

def extract_features(image, kernels):
    feats = np.zeros((image.shape[0],image.shape[1],kernels.shape[2]), dtype=np.double)
    #print(feats.shape)
    for k in  range(kernels.shape[2]):
        filtered = ndi.convolve(image, kernels[:,:,k], mode='wrap')
        feats[:,:,k]=filtered
    return feats



image1=np.asarray(a_sample[0][0])
feature_map = extract_features(image1, F)
H, W, C = feature_map.shape
#print(feature_map.shape)
# 将特征图转换成二维矩阵，形状为 (H*W, C)
flattened_map =  np.reshape(feature_map, (H*W, C))
#print(flattened_map)
# 进行K-means聚类
k=20
kmeans = KMeans(n_clusters=k)  # k是你想要的聚类数量
kmeans.fit(flattened_map)

# 获取每个像素的聚类标签
labels = kmeans.labels_

# 将聚类结果转换回与特征图相同的形状
clustered_map = labels.reshape(H, W)

# 输出聚类结果
print(clustered_map)



# 假设 data 是形状为 (N, 2) 的特征矩阵，N 是样本数量，2 是特征数量
data = flattened_map


# 可视化聚类结果
plt.scatter(data[:, 0], data[:, 1], c=labels, cmap='viridis')
plt.xlabel('Feature 1')
plt.ylabel('Feature 2')
plt.title('Clustering Result')
plt.colorbar()
plt.show()