objdetect: fix out-of-bounds write in HOG gaussian weights (blockSize.height vs width)

In HOGCache::init the block gaussian weights are computed into two buffers:
  AutoBuffer<float> di(blockSize.height), dj(blockSize.width);
The vectorized (CV_SIMD128) loop that fills dj was bounded by
`blockSize.height` instead of `blockSize.width`. When the block is taller than
it is wide, `v_store(_dj + j, ...)` writes past the end of the width-sized dj
buffer, causing a heap buffer overflow / access violation (e.g. HOGDescriptor
with a 198x281 window, or the 2399x2400 case from the report).

Bound the dj loop by `blockSize.width` to match the buffer size and the scalar
tail loop. This keeps the SIMD path (unlike the workaround in the report, which
disabled it).

Added a regression test that computes a HOGDescriptor with a tall block; without
the fix it triggers a heap-buffer-overflow (confirmed with AddressSanitizer at
hog.cpp:731 inside HOGCache::init).

Fixes #23580

modules/objdetect/src/hog.cpp

@@ -723,7 +723,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
     #if CV_SIMD128
         idx = v_float32x4(0.0f, 1.0f, 2.0f, 3.0f);
 
-        for (; j <= blockSize.height - 4; j += 4)
+        for (; j <= blockSize.width - 4; j += 4)
         {
             v_float32x4 t = v_sub(idx, _bw);
             t = v_mul(t, t);

modules/objdetect/test/test_cascadeandhog.cpp

@@ -1364,4 +1364,22 @@ TEST(Objdetect_CascadeDetector, small_img)
     }
 }
 
+// See https://github.com/opencv/opencv/issues/23580
+// HOGDescriptor::compute() overflowed the gaussian weights buffer when the block
+// was taller than it was wide: the SIMD store loop for the width buffer (_dj) was
+// bounded by blockSize.height instead of blockSize.width. The block is chosen wide
+// enough that the buffer is heap allocated, so the overflow is a real out-of-bounds
+// write. "Done" is simply that compute() runs without crashing.
+TEST(Objdetect_HOGDescriptor, issue_23580_tall_block_no_overflow)
+{
+    Size winSize(272, 2048); // height > width, width > AutoBuffer stack size
+    HOGDescriptor hog(winSize, /*blockSize*/ winSize, /*blockStride*/ Size(8, 8),
+                      /*cellSize*/ Size(8, 8), /*nbins*/ 9);
+
+    Mat src(winSize, CV_8UC1, Scalar::all(0));
+    std::vector<float> descriptors;
+    ASSERT_NO_THROW(hog.compute(src, descriptors));
+    EXPECT_FALSE(descriptors.empty());
+}
+
 }} // namespace