Python - cv2.resize with INTER_NEAREST return corrupted slice for channels > 4 #15075
Description
System information (version)
OpenCV => 4.1.0
Operating System / Platform => Ubuntu 16.04.1, Linux 4.13.0
Python => 3.6.8
numpy => 1.16.4
Detailed description
cv2.resize with an array having a 3rd-dimension > 4 (must be an odd number) and nearest-neighbor interpolation returns corrupted data in the last channel. It is not deterministic and looks like memory corruption.
Steps to reproduce
import numpy as np
import cv2
# up to DIM=4 is OK
DIM = 5
T = np.zeros((12, 12,DIM)).astype(np.uint8)
T[2, 2, DIM-1] = 1
print("Sum: ", T.sum())
print(T[:, :, DIM-1])
R = cv2.resize(T, None, fx=0.9, fy=1,interpolation=cv2.INTER_NEAREST)
# it does not help to supply a pre-allocated array either
#R = np.zeros((12, 11, DIM)).astype(np.uint8)
#R = cv2.resize(T, dsize=R.shape[:2], dst=R, interpolation=cv2.INTER_NEAREST)
print("Sum: ", R.sum())
print(R[:, :, DIM-1])Output:
Sum: 1
[[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 1 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]
[0 0 0 0 0 0 0 0 0 0 0 0]]
Sum: 836
[[248 113 0 214 176 85 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]
[ 0 0 0 0 0 0 0 0 0 0 0]]