自研OpenCV warpAffine平移实现与官方结果不符，求排查问题

因业务需求，需基于GPU重实现OpenCV的warpAffine，且仅需支持图像平移操作，对应的2×3变换矩阵形式为：

[[ 1, 0, x_shift],
 [ 0, 1, y_shift]]

为验证算法正确性，先编写了Python原型实现，核心代码如下：

import math
import numpy as np

# Bilinear Interpolation:
# for a source image, sample a value indexed by (y, x) even if x, y are fractionals.
def sample(img, x: float, y: float):
    h, w = img.shape

    left = math.floor(x)
    top = math.floor(y)
    right = left + 1
    bottom = top + 1

    # top left corner
    if (0 <= left < w) and (0 <= top < h):
        a = img[top, left]
    else:
        a = 0

    # top right corner
    if (0 <= right < w) and (0 <= top < h):
        b = img[top, right]
    else:
        b = 0

    # bottom left corner
    if (0 <= left < w) and (0 <= bottom < h):
        c = img[bottom, left]
    else:
        c = 0

    # bottom right corner
    if (0 <= right < w) and (0 <= bottom < h):
        d = img[bottom, right]
    else:
        d = 0

    # linear interpolation of top two points
    top_interleaved = (right - x) * a + (x - left) * b

    # linear interpolation of bottom two points
    bottom_interleaved = (right - x) * c + (x - left) * d

    # linear interpolation of top and bottom points
    return (bottom - y) * top_interleaved + (y - top) * bottom_interleaved


def warpAffine(img, shift_x: float, shift_y: float):
    output = np.empty_like(img)

    h, w = img.shape
    for y in range(h):
        for x in range(w):
            output[y, x] = sample(img, x - shift_x, y - shift_y)

    return output

测试用例与自研实现输出 #

a = np.arange(1, 5, dtype=np.float32).reshape(2, 2)
# a = [[1, 2],
#      [3, 4]]

shift_x = 0.1
shift_y = 0.1

warpAffine(a, shift_x, shift_y)

输出结果：

array([[0.81, 1.71],
       [2.52, 3.7 ]], dtype=float32)

OpenCV官方测试与输出 #

import cv2 as cv

affine_arr = np.array([[1, 0, shift_x], [0, 1, shift_y]], dtype=np.float32)
affine_img = cv.warpAffine(
    a,
    affine_arr,
    (a.shape[1], a.shape[0]),
    flags=cv.INTER_LINEAR,
    borderMode=cv.BORDER_CONSTANT,
    borderValue=0,
)

输出结果：

array([[0.82, 1.73],
       [2.55, 3.72]], dtype=np.float32)

已研读OpenCV warpAffine源码但仍未能定位问题，恳请熟悉该函数的开发者帮忙排查自研实现的问题。

内容的提问来源于stack exchange，提问作者Shuai