From 029d0bef08df49264995c2061e082610f39babe1 Mon Sep 17 00:00:00 2001
From: Milan Nikolic <gen2brain@gmail.com>
Date: Tue, 30 Jun 2026 16:13:32 +0200
Subject: [PATCH] Skip leading bytes before JPEG SOI marker, issue #58

---
 image/decode.go      | 13 ++++++-
 image/decode_test.go | 82 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 94 insertions(+), 1 deletion(-)
 create mode 100644 image/decode_test.go

diff --git a/image/decode.go b/image/decode.go
index 7261fcb..09f856a 100644
--- a/image/decode.go
+++ b/image/decode.go
@@ -4,6 +4,7 @@
 package image
 
 import (
+	"bytes"
 	"image"
 	"io"
 
@@ -22,5 +23,15 @@ type Decoder struct {
 
 // Decode decodes image from JPEG.
 func (d Decoder) Decode() (image.Image, error) {
-	return jpegn.Decode(d.r)
+	data, err := io.ReadAll(d.r)
+	if err != nil {
+		return nil, err
+	}
+
+	// Some cameras emit MJPEG frames with leading bytes before the SOI marker.
+	if i := bytes.Index(data, []byte{0xFF, 0xD8}); i > 0 {
+		data = data[i:]
+	}
+
+	return jpegn.Decode(bytes.NewReader(data))
 }
diff --git a/image/decode_test.go b/image/decode_test.go
new file mode 100644
index 0000000..0381b8b
--- /dev/null
+++ b/image/decode_test.go
@@ -0,0 +1,82 @@
+//go:build !libjpeg
+
+package image_test
+
+import (
+	"bytes"
+	"image/jpeg"
+	"testing"
+
+	"github.com/gen2brain/jpegn"
+
+	"github.com/gen2brain/cam2ip/image"
+)
+
+// stripDHT removes the Huffman table segments, like webcams that rely on the standard tables (issue #20).
+func stripDHT(data []byte) []byte {
+	out := append([]byte{}, data[0], data[1])
+
+	for i := 2; i+4 <= len(data); {
+		if data[i] != 0xFF {
+			break
+		}
+
+		marker := data[i+1]
+		if marker == 0xDA {
+			return append(out, data[i:]...)
+		}
+
+		length := int(data[i+2])<<8 | int(data[i+3])
+		if marker != 0xC4 {
+			out = append(out, data[i:i+2+length]...)
+		}
+
+		i += 2 + length
+	}
+
+	return out
+}
+
+func TestDecodeMissingDHT(t *testing.T) {
+	src, err := jpeg.Decode(bytes.NewReader(testJpg))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// stdlib uses the standard tables; strip the DHT so the stream relies on the decoder's defaults.
+	var buf bytes.Buffer
+	if err := jpeg.Encode(&buf, src, &jpeg.Options{Quality: 90}); err != nil {
+		t.Fatal(err)
+	}
+	data := stripDHT(buf.Bytes())
+
+	if _, err := jpeg.Decode(bytes.NewReader(data)); err == nil {
+		t.Log("note: stdlib decoded a DHT-less stream")
+	}
+
+	img, err := image.NewDecoder(bytes.NewReader(data)).Decode()
+	if err != nil {
+		t.Fatalf("decode without DHT: %v", err)
+	}
+
+	if img.Bounds() != src.Bounds() {
+		t.Fatalf("bounds = %v, want %v", img.Bounds(), src.Bounds())
+	}
+}
+
+func TestDecodeLeadingGarbage(t *testing.T) {
+	data := append([]byte{0x00, 0x00, 0xFF, 0x12, 0x34}, testJpg...)
+
+	if _, err := jpegn.Decode(bytes.NewReader(data)); err == nil {
+		t.Log("note: jpegn decoded leading garbage directly")
+	}
+
+	img, err := image.NewDecoder(bytes.NewReader(data)).Decode()
+	if err != nil {
+		t.Fatalf("decode with leading garbage: %v", err)
+	}
+
+	if img.Bounds().Dx() == 0 || img.Bounds().Dy() == 0 {
+		t.Fatal("empty image")
+	}
+}