其中，src是輸入的圖像矩陣；dst是輸出的圖像矩陣；flags是一個(gè)可選參數(shù)，用于指定變換的類型，其默認(rèn)值是cv2.DCT_II。

二、理解cv2.dct的幾個(gè)變換類型

cv2.dct支持多種變換類型，主要有：

三、應(yīng)用cv2.dct進(jìn)行圖像壓縮和特征提取

cv2.dct通常用于圖像壓縮和特征提取。在圖像壓縮中，我們使用DCT變換將時(shí)域信號(hào)轉(zhuǎn)換為頻域信號(hào)，然后舍棄高頻信號(hào)（通常是小于20%的高頻分量），最后使用DCT-III將頻域信號(hào)轉(zhuǎn)換回時(shí)域信號(hào)。通過這種方式，我們可以有效地壓縮圖像，并減少文件的大小。

import cv2
import numpy as np

# read input image
img = cv2.imread('input.png')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# apply discrete cosine transform
dct = cv2.dct(np.float32(gray)/255.0)

# threshold coefficients
thresh = 0.2
dct_thresh = dct * (np.abs(dct) > (thresh*np.amax(dct)))

# apply inverse discrete cosine transform
idct = cv2.idct(dct_thresh)*255.0

# display output
cv2.imshow('Input', gray)
cv2.imshow('Compressed', np.uint8(idct))
cv2.waitKey(0)
cv2.destroyAllWindows()

import cv2
import numpy as np

# read input image
img = cv2.imread('input.png')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# apply DCT to input image
dct = cv2.dct(np.float32(gray)/255.0)

# extract features
thresh = 0.1
dct_thresh = dct * (np.abs(dct) > (thresh*np.amax(dct)))

# apply inverse DCT to reconstructed image
reconstruction = cv2.idct(dct_thresh)

# display output
cv2.imshow('Input', gray)
cv2.imshow('Features', np.uint8(dct_thresh))
cv2.imshow('Reconstruction', np.uint8(reconstruction))
cv2.waitKey(0)
cv2.destroyAllWindows()