diff --git a/internal/infrastructure/docparser/mineru_cloud_converter.go b/internal/infrastructure/docparser/mineru_cloud_converter.go
index 133d2aec05..a82b9cbdab 100644
--- a/internal/infrastructure/docparser/mineru_cloud_converter.go
+++ b/internal/infrastructure/docparser/mineru_cloud_converter.go
@@ -350,7 +350,10 @@ func (c *MinerUCloudReader) extractDoneResult(_ context.Context, item *extractRe
 
 // --- ZIP handling ---
 
-var imgRefPattern = regexp.MustCompile(`!\[[^\]]*\]\(([^)]+)\)`)
+var (
+	imgRefPattern     = regexp.MustCompile(`!\[[^\]]*\]\(([^)]+)\)`)
+	htmlImgRefPattern = regexp.MustCompile(`(?i)<img\b[^>]*\ssrc\s*=\s*(?:"([^"]+)"|'([^']+)')`)
+)
 
 func downloadAndExtractZip(zipURL string) (string, []types.ImageRef, error) {
 	if err := utils.ValidateURLForSSRF(zipURL); err != nil {
@@ -406,8 +409,19 @@ func downloadAndExtractZip(zipURL string) (string, []types.ImageRef, error) {
 	// Extract referenced images
 	var imageRefs []types.ImageRef
 	seen := map[string]bool{}
+	var referencedPaths []string
 	for _, match := range imgRefPattern.FindAllStringSubmatch(mdText, -1) {
+		referencedPaths = append(referencedPaths, match[1])
+	}
+	for _, match := range htmlImgRefPattern.FindAllStringSubmatch(mdText, -1) {
 		imgPath := match[1]
+		if imgPath == "" {
+			imgPath = match[2]
+		}
+		referencedPaths = append(referencedPaths, imgPath)
+	}
+
+	for _, imgPath := range referencedPaths {
 		if strings.HasPrefix(imgPath, "http://") || strings.HasPrefix(imgPath, "https://") || strings.HasPrefix(imgPath, "data:") {
 			continue
 		}
diff --git a/internal/infrastructure/docparser/mineru_cloud_converter_test.go b/internal/infrastructure/docparser/mineru_cloud_converter_test.go
new file mode 100644
index 0000000000..974f306f87
--- /dev/null
+++ b/internal/infrastructure/docparser/mineru_cloud_converter_test.go
@@ -0,0 +1,70 @@
+package docparser
+
+import (
+	"archive/zip"
+	"bytes"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/Tencent/WeKnora/internal/utils"
+)
+
+func TestDownloadAndExtractZipExtractsHTMLImageReferences(t *testing.T) {
+	utils.SetSSRFWhitelistFromRaw("127.0.0.1")
+	t.Cleanup(func() {
+		utils.SetSSRFWhitelistFromRaw("")
+	})
+
+	zipData := buildMinerUCloudTestZip(t, map[string][]byte{
+		"full.md":            []byte(`<table><tr><td><img src="images/diagram.jpg" alt="diagram"/></td></tr></table>`),
+		"images/diagram.jpg": []byte("jpeg bytes"),
+	})
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, _ *http.Request) {
+		w.Header().Set("Content-Type", "application/zip")
+		_, _ = w.Write(zipData)
+	}))
+	t.Cleanup(server.Close)
+
+	_, imageRefs, err := downloadAndExtractZip(server.URL)
+	if err != nil {
+		t.Fatalf("downloadAndExtractZip() error: %v", err)
+	}
+	if len(imageRefs) != 1 {
+		t.Fatalf("image refs = %d, want 1", len(imageRefs))
+	}
+
+	ref := imageRefs[0]
+	if ref.OriginalRef != "images/diagram.jpg" {
+		t.Errorf("OriginalRef = %q, want %q", ref.OriginalRef, "images/diagram.jpg")
+	}
+	if ref.Filename != "diagram.jpg" {
+		t.Errorf("Filename = %q, want %q", ref.Filename, "diagram.jpg")
+	}
+	if ref.MimeType != "image/jpeg" {
+		t.Errorf("MimeType = %q, want %q", ref.MimeType, "image/jpeg")
+	}
+	if !bytes.Equal(ref.ImageData, []byte("jpeg bytes")) {
+		t.Errorf("ImageData = %q, want %q", ref.ImageData, "jpeg bytes")
+	}
+}
+
+func buildMinerUCloudTestZip(t *testing.T, files map[string][]byte) []byte {
+	t.Helper()
+
+	var buf bytes.Buffer
+	writer := zip.NewWriter(&buf)
+	for name, data := range files {
+		file, err := writer.Create(name)
+		if err != nil {
+			t.Fatalf("create zip entry %q: %v", name, err)
+		}
+		if _, err := file.Write(data); err != nil {
+			t.Fatalf("write zip entry %q: %v", name, err)
+		}
+	}
+	if err := writer.Close(); err != nil {
+		t.Fatalf("close zip: %v", err)
+	}
+	return buf.Bytes()
+}